Spatial variability of microzooplankton grazing on phytoplankton in coastal southern Florida, USA

Sample sites

Measurements were made aboard the R/V Walton Smith at five stations (Fig. 1) from January 2018 through January 2020 during South Florida Ecosystem Restoration cruises that occurred every other month. This consisted of three stations along the Florida reef tract—Molasses Reef (MR—25°0′31″N, 80°22′38″W), 21/Looe Key (LK—24°32′08″N, 81°24′74″W), and Western Sambo (WS—24°28′54″N, 81°43′12″W)—and two offshore of the Shark River outflow into the Gulf of Mexico—54 (25°20′42″N, 81°9′43″W) and 57 (25°21′10″N, 80°16′02″W).

Environmental conditions

Typically, these cruises collect environmental data (temperature, salinity, nutrient and chlorophyll a concentrations, etc.) at up to 118 stations. Environmental data were collected using a Sea Bird 911plus Conductivity-Temperature-Depth (CTD) system mounted on a rosette with 12 Teflon-coated 10-L Niskin bottles. Temperature was measured directly from the CTD and salinity was calculated from the conductivity and temperature measurements. The Niskin bottles collected water just below the surface to measure soluble reactive phosphorus (SRP), ammonia, nitrate + nitrite (NO_x), and silica concentrations.

Nutrient samples were filtered through a 0.45 µm filter into polystyrene test tubes. The samples were frozen and stored at –20 °C until analysis at NOAA’s Atlantic Oceanographic and Meteorological Laboratory (AOML). Nutrients were measured on a SEAL Analytical autoanalyzer using standard gas-segment continuous flow colorimetric methods (Gordon et al., 1993). NO_x was measured using a copper-coated cadmium reduction column to reduce nitrate to nitrite. Then, the nitrite was diazotized with sulfanilamide and coupled with N-1-naphthylethylenediamine dihydrochloride to form an azo dye with its absorbance measured at 540 nm (Zhang, Ortner & Fischer, 1997). Nitrate concentrations were obtained by subtracting nitrite values, which were measured separately using the previous procedure without the cadmium reduction, from the NO_x values (Zhang, Ortner & Fischer, 1997). Ammonia was measured via reaction with alkaline phenol and dichloroisocyanuric acid sodium salt (NaDTT) in the presence of sodium nitroferricyanide as a catalyst at 60 °C to form indophenol blue with its absorbance measured at 640 nm (Zhang et al., 1997). To measure SRP, hydrazine was used to reduce 12-molybdophosphoric acid to phosphomolybdenum blue with the absorbance measured at 880 nm (Zhang , Fischer & Ortner, 2000). Silicate was measured via reaction with molybdate in acidic solution to form β-molybdosilicic acid. The β-molybdosilicic acid was reduced using ascorbic acid to form molybdenum blue with the absorbance measured at 660 nm (Zhang & Berberian, 1997).

Primary production

Primary productivity was estimated at the same five stations mentioned previously via changes in dissolved oxygen (O₂) concentrations over 24 h using the light-dark bottle method incubated in a continuous-flow water bath with ambient seawater on the deck of the research vessel (Cullen, 2001; Wielgat-Rychert et al., 2017). Water for the experiments was collected from just below the surface using the Niskin bottles on the CTD, then gently syphoned into 2 L Polycarbonate Nalgene bottles and sealed by rubber stoppers to avoid atmospheric oxygen introduction. Light bottles were covered in one layer of mesh that reduce the ambient light by ~35%, in order to reduce the chance of photoinhibition (Staehr et al., 2016). Triplicate samples were collected in 125 mL Erlenmeyer flasks at T₀ and T₂₄ from the light and dark bottles. Samples were preserved with manganous chloride (MnCl₂) immediately followed by the addition of an alkaline sodium hydroxide-sodium iodide solution (NaOH/NaI) and stored in the dark with water level maintained around the glass stopper to minimize oxygen transfer. These samples were then analyzed in the AOML laboratory by Winkler titration using the amperometric technique to estimate the dissolved O₂ concentration (Carpenter, 1995; Langdon, 2010).

Net primary productivity (NPP) was calculated by subtracting the initial O₂ concentration from the final O₂ concentration in the light bottles. Respiration (R) was calculated by subtracting the initial O₂ concentration from the final O₂ concentration in the dark bottles. Gross primary productivity (GPP) was calculated by subtracting R from NPP (Carritt & Carpenter, 1966; Duarte, Agustí & Regaudie-de-Gioux, 2011; Sanz-Martín et al., 2019). The change in O₂ concentration was converted to carbon, assuming a photosynthetic quotient of 1.28, which is consistent with recent estimates and larger meta-analyses, and a respiratory quotient of 0.89 (Laws, 1991; Wielgat-Rychert et al., 2017). The change in carbon concentration (µM C) over time was converted to change in biomass per unit volume over time to estimate primary production (gC m^–3 day^–1) in the mass of water collected for all experiments.

Duplicate sub-samples of 200 mL were collected to estimate chlorophyll a concentrations at T₀ and T₂₄ from the light bottles. This was used to calculate the change in phytoplankton biomass over the primary production experiments. The 200 mL water samples were filtered onto GF/F filters and stored in liquid nitrogen on the ship and in a –80 °C freezer upon return to the laboratory. Chlorophyll a was extracted and measured within 3 months of collection. The extractions were made in a 60:40 Acetone:Dimethyl Sulfoxide solution and measured on a TD-700 fluorometer following the procedures of Shoaf & Lium (1976). The extractions were then acidified with two drops of 10% HCl and the acidified fluorescence recorded to correct for phaeophytin. The fluorescence values were calibrated using known concentrations of chlorophyll a to yield chlorophyll a concentration in µg L^–1 (Kelble et al., 2005). Chlorophyll a concentrations in the T₀ bottles were used as the estimation of in situ concentrations in the surface water.

Microzooplankton grazing and phytoplankton growth

In addition to all the work normally conducted on these 5-day cruises, experiments were set-up to measure microzooplankton grazing and phytoplankton growth at the same stations where primary production data has been collected since 2014 (Fig. 1). Dilution experiments, as described by Landry & Hassett (1982), were used to measure the community microzooplankton grazing coefficients (Chl a Chl a^–1 d^–1) on chlorophyll a concentrations (phytoplankton biomass) and the instantaneous phytoplankton growth rate (Chl a Chl a^–1 d^–1). Water for the experiments were collected just below the surface using the Niskin bottles on the CTD. All water was filtered through a 200 µm mesh to remove all grazers larger than microzooplankton, referred to as whole water hereafter. Filtered water used in the dilution experiments was made using a 0.2 µm Life Sciences pleated capsule filter. The screened 2 L Nalgene bottle from the primary production experiments (described previously) was used as the 100% whole water treatment. Triplicate 1 L Corning culture flasks were used for each of 20%, 10%, and 5% whole water treatments. The whole and filtered water for each treatment were combined in a 20 L carboy, and then gently transferred to four 1 L flasks (1 initial + 3 final). None of our treatment bottles were amended with inorganic nutrients. The initial flask was used to collect T₀ data. The final flasks were placed into 1-layer mesh bags and then incubated on deck in a flow through chamber for 24 h.

Water was collected from all the dilution flasks at the start (T₀) and end (T₂₄) of our experiments; one sample from the initial treatment flask and one sample from each triplicate T₂₄ flasks. Samples were analyzed for chlorophyll a as stated above. For each treatment, we used the T₀ and T₂₄ values of chlorophyll a concentrations to calculate the apparent growth rate of phytoplankton biomass using the exponential growth equation (Landry & Hassett, 1982). For each experiment, a linear regression was fit to the apparent growth rates for the 10 treatment bottles (100% (1), 20% (3), 10% (3), 5% (3)). The slope of the line (m) was the microzooplankton community grazing coefficient (g) and the y-intercept (y) was the phytoplankton instantaneous growth rate (µ) (Landry & Hassett, 1982).

We visually assessed the fit of each linear regression and divided the results into four categories; a standard, zero, negative, and non-linear response. A standard response refers to experiments with the expected results, that phytoplankton growth increases as the proportion of whole water decreases (Fig. S1A). The fit of the linear regression for the standard response can vary based on how well all data points fall along the fitted line, but the general pattern is observable. A zero response refers to experiments where minimal difference between the phytoplankton growth rate at 100% whole water and most dilution treatments were observed (Fig. S1B). A negative response refers to experiments where the phytoplankton growth rate was highest in the 100% whole water treatment and decreased in dilution treatments (Fig. S1C). A non-linear response refers to experiments where the phytoplankton growth rate did not substantially differ from the 100% whole water treatment until the 10% or 5% whole water treatments (Fig. S1D). Regardless of the observed responses, we used a fitted linear regression to obtain our measurements of µ and g in order maintain consistency across all experiments. We calculated the daily accumulation rate by subtracting µ from g and the percent of production consumed by dividing g over µ and multiplying it by 100 (Calbet & Landry, 2004; Anderson & Harvey, 2019).

Statistical analysis

The linear regression analysis for all dilution experiments was conducted in Excel using the Analysis ToolPak. Kruskal-Wallis and associated Dunn test analyses to assess differences between stations were conducted in R (version 3.6.3) using the “dunn.test” package (Dinno, 2017). These are non-parametric tests because some of our environmental data failed a homogeneity of variance test. We performed a PCA with temperature, salinity, and chlorophyll a, NO_x, NH₃, phosphate, and silicate concentrations collected on each sample data and location to assess the similarity in the environment between each station. The PCA was conducted in R using the built-in function “prcomp”. Data from the PCA were extracted and visualized using the “factoextra” package for R, including concentration ellipses with 95% confidence intervals (Kassambara & Mundt, 2020).

Environmental data

There were numerous differences in environmental factors between the five stations (Table 1), primarily between the Florida Keys (FK—MR, 21/LK, and WS) and southwest Florida inner shelf (SWF—57 and 54). Average chlorophyll a and silicate concentrations were significantly higher at stations 57 and 54 compared to stations MR, 21/LK, and WS. Average salinity at station 54 was significantly lower compared to all stations, while station 57 was significantly lower compared to MR, 21/LK, and WS (Table 1). There were no significant differences in average temperatures, DIN:DIP, and NO_x, ammonia and phosphate concentrations. However, station 57 and 54 had noticeability higher concentrations of all nutrients but high variability between sampling dates kept these values from being significantly higher. Stations MR, LK/21, and WS in the coastal region of Florida Keys grouped tightly together and stations 57 and 54 in the SWF inner shelf grouped together (Fig. 2). In all ensuing analyses, these two groupings were separated.

Primary production experiments

GPP was highest at station 54 followed by station 57, with averages (±SE) of 0.35 ± 0.08 gC m^–3 d^–1 and 0.17 ± 0.03 gC m^–3 d^–1 produced, respectively (Fig. 3). GPP rates at both of these stations were significantly higher than 21/LK and WS, and station 54 was significantly higher than 57. Respiration was highest at stations 57, 54, and MR, with averages (±SE) of –0.16 ± 0.04 gC m^–3 d^–1, –0.14 ± 0.05 gC m^–3 d^–1, and –0.14 ± 0.09 gC m^–3 d^–1 consumed, respectively (Fig. 3). However, rates of respiration were not significantly different between stations. NPP values suggested that, on the whole, the amount of carbon produced and consumed through respiration every day were in balance, with the exception of station 54 (Fig. 3). Average (±SE) NPP at station 54 (0.21 ± 0.07 gC m^–3 d^–1) was significantly higher compared to stations MR and 57, suggesting that more carbon was being produced than consumed through respiration.

Microzooplankton grazing

Of the 55 dilution experiments we conducted, three had negative microzooplankton grazing rates and 14 had positive grazing rates that were not significantly different from zero (Tables S1 and S2). In the coastal Florida Keys region, µ and g appeared to be in balance at stations MR and LK/21 (Figs. 4A and 4C). The average (±SE) accumulation rates at stations MR (–0.03 ± 0.16 d^–1) and 21/LK (0.02 ± 0.15 d^–1) were essentially zero, but there was high variability in the accumulation rates between sampling dates (Figs. 4A and 4C). At station WS, the average µ (0.98 ± 0.12) was significantly higher (paired t-test, P = 0.003) than the average g (0.60 ± 0.09). Station WS was the only station with an average accumulate rate (0.38 ± 0.10 d^–1) above zero (Fig. 4C). At all three stations in the coastal region of Florida Keys, there was no linear relationship between g and µ (simple linear regression, P-value > 0.05, not shown).

In the SWF inner shelf region, microzooplankton grazing rates tended to outpace instantaneous phytoplankton growth rates at both stations (Figs. 4B and 4D). The average microzooplankton grazing rates (0.90 ± 0.21) at station 57 was significantly higher (paired t-test, P = 0.04) than the average instantaneous phytoplankton growth rate (0.66 ± 0.17) and the average microzooplankton grazing rates (1.21 ± 0.30) at station 54 was also significantly higher (paired t-test, P = 0.02) than the average instantaneous phytoplankton growth rate (0.93 ± 0.23). As a result, the average accumulation rates at stations 57 (–0.24 ± 0.13 d^–1) and 54 (–0.28 ± 0.11 d^–1) were below zero (Fig. 4D). At both stations, there was a significantly linear relationship with g and µ, suggesting that µ and g increase together (simple linear regression, P-value < 0.05, not shown).